Dresdenboys' blog: AMD Bulldozer - Patent based research

[savantu]

My argument is based off the diagrams we've seen so far, which all point to sizable gains. AMD themselves said 35%, I personally think it'll be much closer to 20% on average, which is still a really big increase. The rest of the math was done based off of logic.

When you have 33% more cores, you better get 20% on average. Where is the miracle though?

You can't do math can you? All the evidence so far has pointed to fairly good gains in IPC (just look at Dresdenboy's work), there is literally no logical reason to think that it will be slower.

Ah the irony...Please tell him his math is also wrong.

I think a big problem in all these discussions about per core performance is, that looking at the whole processor's performance is not the best way to guess single core performance. With a hypothetical perfect scaling a 12 core BD based processor (the name of the core actually is "Orochi", BD was just the project according to JF), the per core performance could be higher than for MC, while the single core performance could be lower - because MC needs a higher single core performance to offset losses due to scaling.

When your 2 cents, include Nehalem being "at least" 50% faster then C2D, and being the biggest achievement ever in CPU's history, they really don't mean much.

Are you quoting me on those ? If yes, I don't remember saying that.

OTOH, answering messages 2 months old deserves some attention and those statements are valid : Nehalem and its 32nm derivate are at least 50% faster in any meaningful app ( and I mean commercial workloads, not games ) than the Core 2 generation. I can back that up with dozens of world records for any given socket number.

As for being the greatest, well, what can I say, I let the market decide :

Around 400 million 45nm Nehalem processors have shipped to date, Perlmutter said during a speech at IDF

http://www.pcworld.com/article/19408...arly_2011.html

It is the biggest achievement at least to date given stellar performance, energy efficiency, scaling and features. Whether the new uarch, Sandy Bridge can expand on this remains to be seen.

[Helmore]

When you have 33% more cores, you better get 20% on average. Where is the miracle though?

According to AnandTech, a 16 core (2 × 8 core, like Magny Cours) is around 60 to 80% faster than a 12 core Magny Cours CPU in SPECInt_rate2006. Now you do the math. That's only Integer performance though, but floating point performance of a single module is rumored to also be faster than a dual core Deneb.

We'll just have to wait until Bulldozer is released before we start to draw any conclusions about its performance though. We don't want this to turn into another Barcelona or R600.

[Oliverda]

Ok, That's enough.
I've PM'd savantu to stay out of the AMD threads.
I'm going to say it one more time: Stop the fanboy crap.
It's like some of you sit in wait for an opportunity to jump on someone.
Let it go or I'll let you go elsewhere.

Oops... looks like somebody doesn't read the PMs.

[Dimitriman]

Oops... looks like somebody doesn't read the PMs.

I say let him talk, will make it all that much worse for him if BD does thrash all Intel lineup.

[vietthanhpro]

Bulldozer :ROFL:

[Helmore]

I say let him talk, will make it all that much worse for him if BD does thrash all Intel lineup.

Don't try to provoke him, what you're doing is a little immature. Nothing personal against you, just saying.

[Dimitriman]

Don't try to provoke him, what you're doing is a little immature. Nothing personal against you, just saying.

What I am doing really? I said the best way that biased opinions can be handled is by being proven wrong.

Also no offense but,

Ironic how you pick on my quote for being immature but blatant Intel fanboyism on AMD news is taken seriously?

[savantu]

Oops... looks like somebody doesn't read the PMs.

Au contraire mon ami; I've read that PM when it was send to me and I followed it to the letter. Staying out of a discussion that got heated is something different in my view, than a ban.

As long as the discussion revolves around facts and is carried out in a civil manner, I feel it is a bit unjust to deny someone the right to participate simply because he has a different view. Well, as long as we behave, we should have freedom of speech even on a forum. From my POV I behaved well within the guidelines; I have never engaged in slander, name calling, middle fingers and so on. I base my posts on facts and logical deduction; I have no problem in supporting my assertions and engaging with others in lengthy debates. If his evidence > mine and his logical reasoning > mine, I can accept it like a man; there's more courage in that than having personal delusions on how the world, a company or a product really is or will be.

This makes me wonder for the motives of those who quickly brought the PM in discussion. What are you afraid of ? Can't stand having your favorite company or product in a discussion with pros/cons ? Can't think of anything of value to add and this causes frustration ? I really would like to know.

Btw, the definition of a forum is "as a gathering place of great social significance, and often the scene of diverse activities, including political discussions and debates, rendezvous, meetings, et cetera."

What I am guilty of and I'm aware of that is a healthy dose of irony and sarcasm. I pity those who aren't showing a flicker of irony and can't take it. Go cry in your corner.

Brain research tells us that only twenty percent of human beings have a sense of irony, which means that eighty percent of the world takes everything at face value.

So excuse me, but I'll carry on responding when others reply to my previous posts.

[MrMojoZ]

Au contraire mon ami; I've read that PM when it was send to me and I followed it to the letter. Staying out of a discussion that got heated is something different in my view, than a ban.

Considering your sig is nothing but attacks, everything you say is just to further your anti-AMD agenda. Every post you make is flame baited, how is that following the PM to the letter?

[savantu]

According to AnandTech, a 16 core (2 × 8 core, like Magny Cours) is around 60 to 80% faster than a 12 core Magny Cours CPU in SPECInt_rate2006. Now you do the math. That's only Integer performance though, but floating point performance of a single module is rumored to also be faster than a dual core Deneb.

Based on the picture above, I'd say it is 30% faster for Int and 50% for FP.
So 33% more cores and xx% more frequency ( as a rule of thumb you should get 20% more due to the process shrink ).

The problem is simple : a MC core has more execution units than a BD core ( half a module ). That's why for Int it could very well go down for single thread but exceed MC in through output , while for FP it boils down to FMAC.

We'll just have to wait until Bulldozer is released before we start to draw any conclusions about its performance though. We don't want this to turn into another Barcelona or R600.

That's not really up to us, is it ? At best we can engage in speculation based on tiny bits of info and see who gets closer when the product is actually released.

[savantu]

Considering your sig is nothing but attacks, everything you say is just to further your anti-AMD agenda. Every post you make is flame baited, how is that following the PM to the letter?

So the problem is my sig? I have an agenda ? That's news to me; I haven't realized for a second that I have an agenda. And if you can't see the humor and artistic portrayal of a period in recent history by fellow member cupholder, than what can I say ? The issues are definitely not on this side of the screen.
How exactly are my posts flame baited ? Or is it so for some because that's all they can see ? Plenty seem to consider my posts just fine and answer them based on arguments and not tinfoil hats.What's flame bait in making assertions about a product and presenting evidence to support my assertion ?

We're derailing this thread, I'm not the subject of discussion here.

[Manicdan]

using a projection chart made a year ago sure does look like you know how to use logic and facts

[savantu]

using a projection chart made a year ago sure does look like you know how to use logic and facts

MC is shown clearly on the chart. Given that BD might have very well taped out in this time frame I don't find it at all unreasonable for those estimates to be in a ballpark range.
Have any better estimates ?

[Manicdan]

how about actual tests, we have x6 and x12 cpus now, why not look for benchmarks?

[informal]

This is all just speculation based on a few patents and a vague AMD chart in which (how conveniently) BD perf. bars fade out in the chart...
I hereby "predict" BD will be quite a bit faster core fore and clock for clock than 10h(45nm iteration). 16 core(8 module) Interlagos will just crush MC in every imaginable aspect,be it single thread perf.,perf./watt,multithread perf.,pure integer throughput,pure fp throughput etc.

[Helmore]

http://www.pcper.com/images/news/bulldozer_perf.jpg

Based on the picture above, I'd say it is 30% faster for Int and 50% for FP.
So 33% more cores and xx% more frequency ( as a rule of thumb you should get 20% more due to the process shrink ).

The problem is simple : a MC core has more execution units than a BD core ( half a module ). That's why for Int it could very well go down for single thread but exceed MC in through output , while for FP it boils down to FMAC.

I was using the numbers that AnandTech provided over on this article of theirs: http://it.anandtech.com/show/2879/3
Look at the 'quote' just before the conclusion.
It's hard to speculate about performance at this point though.

That's not really up to us, is it ? At best we can engage in speculation based on tiny bits of info and see who gets closer when the product is actually released.

I was mainly talking about the hype and the resulting disappointment of both Barcelona and R600. That part is up to us.
Some people over here are already claiming that Bulldozer will completely obliterate what Intel will have by the time Bulldozer is released (alright, I'm exaggerating a bit ). I don't think those who claim that are really serious about their claim though, but some people might take their claim too seriously .

[Dimitriman]

The consequences of AMD failing to at least slightly improve performance over Gulftown and equivalents with Bulldozer are enormous. It has survived the past 5 years under the shadow of Intel with significant financial backing from external sources and had to become fabless in order to not file for bankrupsy.

So I think feeling good or taking any pleasure in predicting that Bulldozer will fail is not thinking straight. No matter how much you love your favorite brand/company, you should never love spending $1000 on just a CPU. And thats the destiny of anyone buying Intel if AMD fails to deliver competition with BD.

[AliG]

When you have 33% more cores, you better get 20% on average. Where is the miracle though?




Ah the irony...Please tell him his math is also wrong.

That's 20% ipc (which means logic per core), learn to do some research before you try flaming people with more experience than you.

I cannot tell, is your sole goal to flaunt your intel-biased thoughts in an obtuse manor or are you legitimately trying to help the thread?

I suggest you think considerably before your next anti-amd tirade of quotes, and ask yourself if you are adding anything to the thread. If the answer is yes, be my guest and continue to post. Otherwise, and you will certainly know the answer better than anyone else, you already know what to do.

[Manicdan]

i would like to see what happens if both companies had a cpu with the same exact performance and all other main features, would they both be 1000$ or like 300$? we havnt seen a point where both high end options were ultra competitive with each other

[god_43]

i thought ipc ment instructions per cycle? wiki agrees with me?

Both simple pipelining and superscalar design increase a CPU's ILP by allowing a single processor to complete execution of instructions at rates surpassing one instruction per cycle (IPC).[16] Most modern CPU designs are at least somewhat superscalar, and nearly all general purpose CPUs designed in the last decade are superscalar. In later years some of the emphasis in designing high-ILP computers has been moved out of the CPU's hardware and into its software interface, or ISA. The strategy of the very long instruction word (VLIW) causes some ILP to become implied directly by the software, reducing the amount of work the CPU must perform to boost ILP and thereby reducing the design's complexity.
[edit] Thread level parallelism

http://en.wikipedia.org/wiki/Central_processing_unit

[savantu]

That's 20% ipc (which means logic per core), learn to do some research before you try flaming people with more experience than you.

It means Instructions per Cycle ; but anyway, nice of you to make a tutorial stance. Where exactly has AMD said anything about single thread performance ? All that I've seen is through output. BD might be a through output monster with similar single thread performance as k10.5; a concept similar to Niagara.
SUN Niagara has less single thread performance than Atom, but on multithreaded tasks it mocks Nehalem.

I cannot tell, is your sole goal to flaunt your intel-biased thoughts in an obtuse manor or are you legitimately trying to help the thread?

I suggest you think considerably before your next anti-amd tirade of quotes, and ask yourself if you are adding anything to the thread. If the answer is yes, be my guest and continue to post. Otherwise, and you will certainly know the answer better than anyone else, you already know what to do.

I'm doing nothing more ,but searching out of curiosity to see how next gen products from Intel and AMD might stack up. My conclusions might be different than yours; discovery goes both ways; I'm sorry if your only mode of "helping a thread" is to pump up the jam.
It's strange how sensitive people are to discussing the pros/cons of a CPU; it's not your family member btw.

[informal]

As for savantu's motives,one can look here or here(romanian forums-note his signatures),or simply look at Ace's HW forum and his posts which are basically wishing - AMD - to fail. Simple google search did the trick great...

[FlawleZ]

As for savantu's motives,one can look here or here(romanian forums-note his signatures),or simply look at Ace's HW forum and his posts which are basically wishing - AMD - to fail. Simple google search did the trick great...

Maybe so, but that would be very foolish on anyone's part, fanboy or not. If AMD were to fail it would likely be a crushing and possible fatal blow to the company as a whole. With AMD out of the picture, you can bet the only things that are are the $$$ in Intel's eyes, and your empty wallet of course.

[Chumbucket843]

Maybe so, but that would be very foolish on anyone's part, fanboy or not. If AMD were to fail it would likely be a crushing and possible fatal blow to the company as a whole. With AMD out of the picture, you can bet the only things that are are the $$$ in Intel's eyes, and your empty wallet of course.

intel would be in violation of the sherman anti-trust act if amd went bankrupt and if amd goes bankrupt they lose their x86 license. intel will be broken up by the supreme court and we might end up with better competition. restructuring would be hard though. they are actually better off where they are right now. they are strong enough to control amd (i.e. sse5, 3dnow, other pc standards)but not enough so that they arent heavily regulated. that's $$$.

[OhNoes!]

Being pro-logic is NOT anti-AMD! We all want AMD to be competitive, but some of you are just brain-washed trolls who are trying to turn XS into the AMDZONE! It's a cult worship if nothing bad can be said about the power you worship!

Unless someone actually has some solid numbers and a trusted source (complete with screenshots) to back them up, then everyone in this thread is speculating. IF that is the case (which it is), why should we all hold hands and sing kumbayah? The irony is you see the same users harassing members in every AMD thread and yet they all get the respect they deserve in every thread. It's pathetic how you guys (you know yourselves) have segmented this forum into AMD and the REST, while actively controlling who has the right to say what and where! Very pathetic. The powers that be really need to open their eyes to the INTOLERANCE. Maybe just go register your own domain and set up a website with a disclaimer that objective/subjective comments about AMD products are not welcome and would result in a swift ban? How about that?

Posting on this website is a privilege, not a right. Those with the AMD cult worship agenda can go to the AMDZONE; the rest of us would discuss the pros and cons of hardware - yeah that includes both AMD and Intel.

/rant